Office Posture Detection Using Ceiling-Mounted Ultra-Wideband Radar and Attention-Based Modality Fusion.

Area of Science:

• Human-Computer Interaction
• Wearable Technology
• Ergonomics

Background:

• Prolonged sedentary behavior in offices increases risks of musculoskeletal and metabolic disorders.
• Existing workplace stretching monitoring methods face privacy and sensor placement limitations.
• Need for effective, privacy-preserving solutions for monitoring office worker postures.

Purpose of the Study:

• To propose and evaluate a ceiling-mounted ultra-wideband (UWB) radar system for privacy-preserving office posture classification.
• To assess the performance of two fusion models, ConcatFusion and AttnFusion, for posture recognition.
• To demonstrate the system's potential for real-time ergonomic assessment and workplace health integration.

Main Methods:

• Collected UWB radar data from ten participants across five scenarios (seated/standing working/stretching, empty environment).
• Transformed radar distance and Doppler information into modality-specific images.
• Utilized ConcatFusion (baseline) and AttnFusion (attention-based) models for posture classification, evaluated via leave-one-subject-out cross-validation.

Main Results:

• The AttnFusion model achieved superior performance compared to ConcatFusion.
• AttnFusion attained a testing accuracy of 90.6% and a macro F1-score of 90.5%.
• Demonstrated high effectiveness of UWB radar and attention-based fusion for posture monitoring.

Conclusions:

• Ceiling-mounted UWB radar offers a viable, privacy-preserving method for office posture monitoring.
• Attention-based modality fusion (AttnFusion) significantly enhances classification accuracy.
• The system shows promise for improving workplace ergonomics and health and safety programs.